A Spatial Downscaling Method for Deriving High-Resolution Downward Shortwave Radiation Data Under All-Sky Conditions

Downward shortwave radiation (DSR) is an essential parameter in land surface energy budget. However, current DSR products are mainly generated at coarse-resolution scales (more than 5 km) and fail to accurately depict DSR distribution over different topographic and land cover conditions. Meanwhile, the existence of frequent cloud cover constrains the high-resolution DSR estimation. To overcome the above issues, a novel spatial downscaling method for high-resolution DSR estimation was proposed in this study by incorporating coarse-resolution Meteosat Second Generation (MSG) DSR product and Landsat-8 observations. Through decomposing the downscaling scheme into three separate models: fully cloudy, partial cloudy, and cloud-free, the 3-km MSG DSR data were spatially downscaled to 30-m scale under all-sky conditions, based on the assumption of scale-invariant of the models established at 3-km scale. An empirical model for DSR estimation under cloud cover condition was constructed between the top of atmosphere radiance from Landsat-8 and MSG DSR. The downscaled results showed reasonable DSR values under different cloud cover conditions, and the spatial heterogeneity of the downscaled DSR was also well depicted with the variation of surface topography. Meanwhile, the validation with in situ measurements also revealed the significant improvement in terms of the coefficient of determination (
$R^{2}$
) (from 0.53 to 0.79) and the root mean squared error (RMSE) (from 198.5 to 140.41 W/
$\text{m}^{2}$
). In general, the proposed downscaling method in this study shows good potential for high-resolution DSR estimation without regard to the atmospheric information required in traditional DSR estimation under all-sky condition.